netdev - Re: [PATCH net v1] ice: do not reserve resources for RDMA when disabled

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <7c63c3db-072c-4f44-a487-f7e7de9f39e4@intel.com>
Date: Fri, 15 Nov 2024 09:51:01 +0100
From: Przemek Kitszel <przemyslaw.kitszel@...el.com>
To: <jbrandeb@...nel.org>
CC: <jbrandeburg@...udflare.com>, <netdev@...r.kernel.org>,
	<intel-wired-lan@...ts.osuosl.org>, Dave Ertman <david.m.ertman@...el.com>,
	Tony Nguyen <anthony.l.nguyen@...el.com>, Andrew Lunn
	<andrew+netdev@...n.ch>, "David S. Miller" <davem@...emloft.net>, "Eric
 Dumazet" <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>, Paolo Abeni
	<pabeni@...hat.com>, "Czapnik, Lukasz" <lukasz.czapnik@...el.com>, "Michal
 Swiatkowski" <michal.swiatkowski@...ux.intel.com>, Pawel Chmielewski
	<pawel.chmielewski@...el.com>
Subject: Re: [PATCH net v1] ice: do not reserve resources for RDMA when
 disabled

On 11/14/24 01:00, jbrandeb@...nel.org wrote:
> From: Jesse Brandeburg <jbrandeb@...nel.org>
> 
> If the CONFIG_INFINIBAND_IRDMA symbol is not enabled as a module or a
> built-in, then don't let the driver reserve resources for RDMA.
> 
> Do this by avoiding enabling the capability when scanning hardware
> capabilities.
> 
> Fixes: d25a0fc41c1f ("ice: Initialize RDMA support")
> CC: Dave Ertman <david.m.ertman@...el.com>
> Signed-off-by: Jesse Brandeburg <jbrandeb@...nel.org>
> ---

Hi Jesse, it's good to hear back from you :)

we are already working on resolving the issue of miss-allocating
too many resources (would be good to know what beyond MSI-x'es
you care about) for RDMA in the default (likely non-RDMA heavy) case. 
Here is a series from Michal that lets user to manage it a bit:
https://lore.kernel.org/netdev/20241114122009.97416-3-michal.swiatkowski@linux.intel.com/T/

and we want to post another one later that changes defaults from the
current "grab a lot when there are many CPUs" policy to more resonable